runIndirect: Run simple indirect meta-analysis for all possible pairwise...

In RichardBirnie/mautils: Utilities to facilitate an automated workflow for meta-analysis

Description

Run simple indirect meta-analysis for all possible pairwise comparisons in a dataset

Usage

runIndirect(df, data_type, direct_results, continuous = FALSE,
  effect_type = "all", back_calc = FALSE, order_treatments = NA)

Arguments

df	A data.frame This should be in one of two formats. Arm level data must contain the columns 'study' and 'treatment' where study is a study id number (1, 2, 3 ...) and treatment is a treatment id number. If the data are binary then the data frame should also contain columns 'responders' and 'sampleSize' reporting the total number of events and total number analysed respectively for each arm. Relative effect data (e.g. log odds ratio, log rate ratio) must contain the same study and treatment columns plus the columns 'diff' and 'std.err'. Set diff=NA for the baseline arm. The column std.err should be the standard error of the relative effect estimate. For trials with more than two arms set std.err as the standard error of the baseline arm. This determines the covariance which is used to adjust for the correlation in multiarm studies. This is used to identify the set of comparisons available in the data set
data_type	A character string specifying which type of data has been provided. Currently only 'treatment difference' or 'binary' are supported
direct_results	A data frame containing the results of direct meta-analysis. These results are required to provide the inputs for the indirect comparisons. The best way to produce this is to use the tools for direct meta-analysis from this package
continuous	Logical (TRUE/FALSE) indicating whether the effect measure is continuous (mean difference) or a ratio measure (odds ratio, hazard ratio etc). This is passed directly to doBucher and then to bucher
effect_type	A character string indicating what kind of analysis is required. Set to 'Fixed' for fixed effect, 'Random' for random effects or 'all' to get both (Default).
back_calc	A logical indicating whether results should be back transformed. This is used to set the corresponding backtransf argument of the underlying functions from the meta package. If backtransf=TRUE then log odds ratios (or hazard ratios etc) will be converted to odds ratios on plots and print outs. Default is FALSE
order_treatments	An optional argument to specify the order in which treatment comparisons are sorted in the output. The default is NA in which case comparisons will be sorted alphabetically by intervention. If a specific order is required then this should be provided as a data frame with two columns named 'description' and 'Order'. Note that column headers are specific and case sensitive. The description column should contain the names of the treatments exactly as they are specified in the data set. The id column should contain the numbered order of treatments required.

Details

This function performs indirect meta-analysis using the Bucher method for all possible comparisons in a given data set. This function takes a set of treatment comparisons from one or more studies and identifies all possible indirect comparisons where two treatments can be connected via a common comparator. If there is more than one way to connect two treatments then all possible variations are calculated. This function calls doBucher internally to calculate the treatment effects

The inputs for this function will usually be the results from direct meta-analysis for a given set of treatments. The recommended workflow is to use the tools from this package to perform head to head meta-analysis for a given set of treatments then use the resulting data frame to provide the inputs for this function.

Value

A data frame containing the results of all possible indirect comparisons in the data set. The help page for doBucher provides a detailed description of the columns in the output

See Also

doBucher

RichardBirnie/mautils documentation built on July 12, 2019, 8:56 p.m.